SP DISCUSSION PAPER NO.9921 20849 OECD Public Pension Programmes in Crisis: An Evaluation of the Reform Options Richard Disney August 1999 ro t6n LABOR MARKETS, PENSIONS SOCIA( ASSISTANCE T HE W OR L D B AN K OECD public pension programmes in crisis: An evaluation of the reform options Richard Disney August 1999 Abstract Public pension programmes in OECD countries are in difficulties. With ageing populations, and declining participation of working age men in paid work, existing pension arrangements are likely to be unsustainable in the future in many of the richer OECD countries. Indeed, supporting existing pension commitments, even before the 'baby boom' generation reaches retirement, has already proved problematic in countries such as Italy. Some governments have already taken steps to tackle the pension issue but there is inevitably conflict over who will bear the burden of retrenchment: will it be current taxpayers, current pensioners, or future generations of taxpayers and pensioners, perhaps not yet born? This paper considers several issues. It examines the evidence as to whether public pension programmes in some richer OECD countries are indeed in need of major surgery, focusing in particular on the issue of fiscal sustainability. It then considers why programmes have got into financial difficulties. Consideration of this issue provides some clues as to what type of reform process is likely to be viable and credible. The paper then examines the strengths and weaknesses of some reform strategies. A central issue considered there is whether pension programmes should be funded or unfunded. OECD public pension programmes in cnsis: An evaluation of the reform options Richard Disney* * Professor of Economics, University of Nottingham, Research Fellow, Institute for Fiscal Studies and Director of Axia Economics. Cormments to Axia Economics, 38 Concanon Road, London SW2 5TA; e-mail: richard.disney@axiaecon.com; web: http:// www.axiaecon.com. Versions of this paper have been presented at ACES at the ASSA meetings in New York, January 1999, and in workshops at the International Monetary Fund and the World Bank. My thanks to Axel B6rsch-Supan, David Greenaway, David Miles, Robert Palacios and Edward Whitehouse for comments. The views expressed here are those of the author alone and the normal disclaimer applies. This paper is part of the World Bank's pension reform primer progranmne. For details see the web site: http://www.worldbank.org/pensions. 1 Table of Contents 1. Introduction .................................................... 3 2. The problem of financing pension schemes in OECD countries ....................... . 4 3. Why have pension liabilities got out of control in so many OECD countries? ..... 11 3.1 The demographic transition .................................................... 11 3.2 The early stages of unfunded pension programmes ............................................... 13 3.3 'Falling labour force participation of older men .................................................... 14 3.4 Forecasting errors ............. 14 3.5 The 1Ponzi Game' nature of unfunded social security ...................................... 16 4. Disparities in private provision .................................... 18 5. A taxonomy of pension reforms ................................... 21 5.1 A 'parametric' reform of the unfunded programme ...................................... 23 5.2 An 'actuarially fair' unfunded programme ...................................... 24 5.3 'Clean-break' privatisation ...................................... 26 5.4 Partial privatisation ...................................... 33 6. Conclusion ........................................ 35 Bibliography ....................................... 36 2 OECD public pension programmes in cnsis: An evaluation of the reform options Richard Disney 1. Introduction Public pension programmes in OECD countries are in difficulties. With ageing populations, and declining participation of working age men in paid work, existing pension arrangements are likely to be unsustainable in the future in many of the richer OECD countries. Indeed, supporting existing pension comnmitments, even before the 'baby boom' generation reaches retirement, has already proved problematic in countries such as Italy. Some governments have already taken steps to tackle the pension issue but there is inevitably conflict over who will bear the burden of retrenchment: will it be current taxpayers, current pensioners, or future generations of taxpayers and pensioners, perhaps not yet born? There are also major differences in philosophy over the nature of pension reforms, both as proposed and implemented. Blueprints exist for what pension schemes should look like, such as the framework provided by the World Bank.' But many OECD countries face substantial trnsiion problems in attempting to move from existing arrangements to what might be optimal in the future and have in practice followed a variety of different paths. An interesting question is whether economic theory gives any guide as to which type of reform will succeed. It may be however that countries are too idiosyncratic in their existing systems and in their political structure as to permit a common approach to reform. This paper considers several issues. It examines the evidence as to whether public pension programmes in some richer OECD countries are indeed in need of major surgery, focusing in particular on the issue of fiscal sustainability. It then considers why programmes See World Bank (1994) and Holzmann (1998a). 3 have got into financial difficulties. Consideration of this issue provides some clues as to what type of reform process is likely to be viable and credible. The paper then exarmines the strengths and weaknesses of some reform strategies. A central issue considered there is whether pension programmes should be funded or unfunded. A aa on coverage of OECD countries should also be noted. After 1994, with the advent of Mexico (1994), the Czech Republic (1995), Hungary, Poland and Korea (1996) as OECD members, the diversity of pension programmes within the organisation has increased substantially. Since the problems of pension systems in transition econormies, and the Latin American and East Asian approaches to pension reform, are somewhat different, these countries are covered separately in the Pension Reform Primer series.2 There is also a lack of published data for Turkey, although its pension schemes in fact share many of the characteristics of other Mediterranean countries. The focus here is therefore on the relatively homogeneous group of richer OECD member states.' 2. The problem of financing pension schemes in OECD countries The problem in the richer OECD countries is largely one of financing puN5k pension programmes: defined as their contribution-based universal scherme of provision of public benefits for the elderly (social security' in US parlance). These public pension programmes are largely uomdaia3 Pension payments are made from current contributions, usually levied as payroUl taxes, but sometimes paid directly out of general revenues.4 Future obligations are conditioned on current rules concerning eligibility and accrual rates and, it is hoped, will be met by future generations of contributors. Prospects for private, largelyfwolt, schemes are considered in the next section. 2 Grandolini and Cerda (1998) on Mexico, Palacios and Rocha (1998) on Hungary, and Chlion, Gora and Rutkowski (1999) on Poland. Surveys of pension reform in Eastern Europe/Central Asia and East Asia are also in preparation. 3 Some of these public programmes are 'partially funded' in that a publicy managed reserve has been built up. See Iglesias and Palacios (1999). 4 A subsidiary debate therefore concerns whether public pension programmes should be financed by an earmarked 'insurance' contribution, so emphasising the 'aauarial basis' of the progrmme, or out of general taxation, implicitly giving weight to the redistributive aspect. There is also the implicit issue of the transparency of a contribution-based scheme relative to a tax-financed scheme. Note that some countnes, such as France and the Netherlands, have recently moved to broaden the base for financing benefits. Others, such as Australia, have always financed public benefits from general revenues. 4 A number of projections have been made by international institutions and individual governments of the future liabilities of public pension programmes and of the consequences of these liabilities for fiscal balances. Such projections rely on assumptions that are not always explicit and, on past evidence, are subject to large errors once outcomes are observed. On the 'over-pessimistic' side, projections often ignore policy feedback mechanisms, which have tended in the past to impose restraints on the growth of liabilities of programmes. On the 'over-optimistic' side, countries have generally been systematically too optimnistic in understating longevity improvements and overstating the labour force participation of older workers. In addition, different projections seem to rank different OECD countries quite differently in terns of the financing 'crises' associated with pension programmes. Table 1 presents some recent projections from OECD itself for a range of countries. Column (2) is a memo item, showing that there is a fair degree of uniformity in ages of entitlement for a full public pension across countries, although there are some interesting outliers such as Japan, Italy and New Zealand on the one hand, and the Scandinavian countries on the other. However it should be noted that official pension ages give litde guide to cross- country variations in retirement behaviour. Many countries, including the United States but not the United Kingdom, have explicit actuarially favourable early retirement options within the public pension programrne. A number of countries bordering the Mediterranean also have or have had devices such as 'seniority pensions', which allow individuals to retire on a full pension after a certain number of years' service, whether or not they have reached the official state pension age.' In contrast, in the countries of northern Europe, retirement through disability benefit schemes and other special early retirement 'windows' (within the public programme) have been popular, especially in periods of recession. The remaining columns are derived from an OECD-based simulation model of pension liabilities for individual countries. The underlying procedure is standard: it projects earnings growth based on an aggregate model of earnings, utilises existing contribution rates and projects benefit 'rules' (taking account of any prospective reforms) in order to calculate S For example, in Italy, untl the recent reforn, 35 years of contributions would automatically entitle an individual to a full pension; however public servants could originally receive a full pension after 20 years' service. In Turkey, 25 years of contribution would entitle the individual to a full pension as would being 55 (a man) or 50 (a woman). In Greece, there is a complicated formula linking days of contributions to first age of receipt, but men could in principle retire at 58 and women at 50. 5 future pension liabilities and contribution receipts. Participation rates are held constant in the OECD case, where it is also assumed that the age-earnings profiles, presumably derived from cross sections, are 'augmented' by 1 12 per cent annual productivity growth and where GDP growth is assumed to be on a 'medium termn growth path'. These assumptions are sufficient to calculate columns (3) and (4), which show the cost of public pensions as a proportion of GDP and the present value of the difference between projected benefits and currently legislated contribution revenues, to the time horizon of 2070.6 It should be noted in passing that there is not a close correlation between the current GDP cost of pension provision and the ratio of the population of 65 and over to the working population (the 'aged dependency ratio' - ADR); see Figure 1. In 1990, for example, the highest dependency ratios, in descending order, among the countries cited here for which there are data were Sweden, Norway, the United Kingdom and Denmark. Yet three of these countries are arnong the lowest in terms of pension payments as a proportion of GDP in 1995. This result is due to a combination of lower benefit levels, higher effective retirement ages and price indexation of pensions. Of course, given the simulation methodology, the increase im public pension payments to 2030 will be much more closely related to demographic change. It is therefore no surpn'se to see countries where the demographic transition to an aged society is dramatic, such as Italy, Japan and the Netherlands, exhibiting rising ratios. It is hard to know what is an acceptable upper limnit on pension payments as a proportion of GDP and it will partly depend on the country-specific demographic transition. If, however, as a rule of thumb, we were to believe that a 10 per cent ratio of public pension payments to GDP should be an effective ceiling on public pension commitments, then some countries already exceed this fraction, and many more wil do so by 2030. 6 This is the tenninal date chosen by OECD. 6 Figure 1. Aged dependency ratio and public pension payments as a percentage of GDP, 1990-95 aged dependency ratio, per cent 30 25- 4 2 0 --+ 204 1 5 10 -_- 0 5 10 15 pension payments, per cent of GDP Column (4) indicates starkly that pension contributions will have to rise if large deficits are not to be incurred. Government actuaries in fact, often automatically project contribution increases in order to ensure that there is no projected deficit. Whether such projected contribution increases are irnplemented in all countries is quite a different matter, and already in several countries existing pension commitments are being partly funded by borrowing. 7 Table 1. OECD Estimates of financial liabilities of public pension programmes 1995-2030 (1) (2) (3) (4) (5) (6) Country Standard age of Public pension Present value of Net financial liabilities, Increase in tax/GDP Entitlement to pension payments, % of GDP contributions less % of GDP ratio required to keep (men/women) 1995 2030 pension spending, 1995 2030 net debt constant % of GDP 2005 2030 United States 65 65 4.1 6.6 -23.0 51 95 -0.3 5.3 Japan 60 58 6.6 13.4 -70.0 11 317 3.5 9.6 Germany 65 65 11.1 16.5 -61.6 44 216 2.8 9.7 France 60 60 10.6 13.5 -102.1 35 165 0.8 7.1 Italy 60 55 13.3 20.3 -59.7 109 234 1.8 11.4 United Kingdom 65 60 4.5 5.5 -23.8 40 137 1.7 3.5 Canada 65 65 5.2 9.0 -100.7 70 -27 -3.2 3.6 Australia 65 60 2.6 3.8 -96.7 28 10 -1.3 2.4 Austria 65 60 8.8 14.4 -92.5 50 317 3.8 15.4 Belgium 65 60 10.4 13.9 -152.6 128 77 -2.0 5.9 Denmark 67 67 6.8 10.9 -234.5 46 34 -1.9 3.8 Finland 65 65 10.1 17.8 -64.8 -7 98 -1.4 8.8 Greece 65 60 .. .. .. .. Iceland 67 67 2.5 4.2 -66.2 35 69 -0.3 4.3 Ireland 66 66 3.6 2.8 -17.8 86 83 -0.3 1.8 Luxembourg 65 65 .. .. Netherlands 65 65 6.0 11.2 -53.3 43 185 0.8 9.0 New Zealand 61 61 5.9 8.3 -212.8 .. Norway 67 67 5.2 10.9 -124.1 -26 -57 -2.7 3.8 Portugal 65 62 7.1 13.0 -109.2 71 170 0.5 8.2 Spain 65 65 10.0 14.1 -108.6 50 159 0.9 7.4 Sweden 65 65 11.8 15.0 -132.3 28 78 -0.6 4.0 Switzerland 65 62 Key. Column 2: standard age of first entitlement to state pension, in 1992 Source: Column 2, 4 and 6: OECD (1996), Tables 2.3 and 5.3; column 3: OECD (1998a); columns 3-6, Roseveare et at. (1996), Tables 1, 3 and 6. Notes: Column 4: the net present value of employer and employee contributions to 2070, net of future pension liabilities and less existing assets, as percentage of 1994 GDP. Assumes annual productivity growth of 1.5 per cent and discount rate of 5 per cent. Column 5: existing revenues and expenditures held constant as a proportion of GDP. Scenario in (3) and (4) for pension expenditure. Health care costs assumed to grow in line with GDP but health care cost-age ratio maintained as at present. Net interest payments are then added and the real interest rate is assumed to reflect (and track) GDP growth. 8 Given the pension calculations, the next stage in such calculations is to take the existing fiscal stance and to factor in the extra pension expenditures to examnine future 'net financial liabilities', so as to project the future overall fiscal balance. These will look larger, of course, if the budget is already in deficit. The calculations in columnn (5) also include the projected costs of health care. They assume that underlying health costs will rise in line with GDP (despite the fact that in most countries they have risen considerably faster in recent years) and adjust le&s of health care costs to the evolving demographic profile (since the very young and the elderly tend to incur more health expenditures than other age groups).' These numbers, accompanied by those in column (6), are those that give politicians and central bankers heart attacks. Projected increases in financial liabilities in countries such as Austria, Japan, Italy and Germany look alarming, and increases in contribution/GDP ratios, to levels as high as 10-15 per cent by 2030, are not seen as electoral vote winners. An important catit, however, is that we have little mneans of judging the reliability of such projections. For example, Canada, although it has recently implemented a pension reform, has a rapidly rising cost of health care and one of the most rapid demographic transitions among these countries, yet it is shown as having an inprovement in its financial position over the period. Vanrations in the range of projections based on alternative assumptions as to demographic and economic variables of course 'fan out' as we project further into the future. As a check on these numbers, Table 2 contains some comparable simulations from the ME. Countries are ranked from the highest net pension liability to the lowest. These figures do not allow for the additional costs of health care. They are calculated on similar assumptions to those of the OECD study, with ll½ per cent productivity growth. Although there appears to be no explicit attempt to model the evolution of wages, the simulations appear to allow for cohort-specific variation in labour force participation rates. The basic message of Table 2 is similar to that of Table 1, but there are important country-specific differences. The increases in debt liabilities are not as dramatic as in the OECD calculations, perhaps because additional health care expenditures are ignored. It is also noticeable that the 7 An important issue is whether increased longevity is associated with better health (i.e. a reduction in morbidity). There is some evidence that age-specific morbidity rates have dedined, but future progress wMil depend on medical advances. 9 IMF method of calculation places a heavy emphasis on cw'ret fiscal stance. For example, Italy is shown to have a positive primary balance in 1995 and thus the net adjustment required appears to be much lower than, say Sweden (in contrast to Table 1, and common sense). Table 2. Net pension liabilities and sustainability of fiscal stance per cent of GDP Net public Net pension Combined net Primary Primary balance needed to: Adjustment in debt liability debt liability balance Stabilise net debt Stabilise long-run Primary balance (1994) (1995-2050) (1995) in 1995 net debt forstability France 42 114 156 -0.3 0.7 4.0 4.3 Germany 53 111 163 2.4 1.1 4,5 2.1 Japan 33 107 140 -0.2 0.3 3.6 3.8 Italy 113 76 188 3.3 2.1 4.6 1.3 Canada 72 68 139 0.2 2.7 4.7 4.5 United States 63 26 89 0.4 1.1 1.9 1.5 Sweden 55 20 75 -5.1 0.1 1.0 6.1 United Kingdom 38 5 42 0.4 0.7 0.8 0.4 Source: Chand and Jaeger (1996), Table 8 A further reservation concerning financial simulations of this type is that they make only lirnited attempts to model the underlying economic structure. Increases in financial liabilities are typically assumed to raise real interest rates, but there is litde attempt to model the impact of this increase on the capital stock. In addition, the underlying determinants of the capital stock and labour productivity may in turn be related to the structure of the working population and the savings rate. The most obvious feedback mechanism is that the age structure affects the aggregate saving rate, which in tum affects the rate of capital accumulation and thus output growth. Models that embody life cycle relationships of this kind have been constructed, but it is generally hard in GE-type model simulations also to calibrate the country-specific complexities of pension accruals and contribution structures across cohorts.' Whatever the limitations of the modelling procedures, Tables 1 and 2 convey a clear picture of growing fiscal imbalances and implicit liabilities that are not usually transparent in 8 For dynamic GE models of this type, see, for example, Auerbach and Kotlikoff (1987). See also Cutler etal (1990). For the United Kingdom in particular see Mles (1998, 1999). 10 national accounts data.9 For some countries, these prospective liabilities look quite alarming. How then did public pension programmes get into this position? Is, it, for example, simply a consequence of the ageing of the OECD population? The next section suggests that this is not the only, or indeed the primary, factor. 3. Why have pension liabilities got out of control in so many OECD countries? 3.1 7he denograpi tsition The standard explanation for the rise in financial liabilities of pension programmes is the ageing of the OECD population. Table Al at the end of the paper provides three measures of dependency for most OECD countries. First, the aged dependency ratio (ADR), as discussed previously. Secondly, the total dependency ratio (MDR), which also includes children aged 0-14. Finally, the 'needs weighted support ratio' suggested by Cutler a al (1990), which takes account of the fact that the resources spent on an elderly person may differ from those spent on a child. All countries, with the exception of Ireland, saw a rise in the ADR between 1960 and 1990 and all countries, including Ireland, exhibit further ageing until 2030. However total dependency ratios, when children are included, generally declined between 1960 and 1990 although they will rise between 1990 and 2030. A mixed picture is observed in terms of the adjusted support ratio (where a decline in the ratio indicates a deterioration). A few countries saw a sharp rise from 1960 to 1990, such as Canada and Iceland, because of the fall in the share of children in the population. However, many see double digit declines in the ratio in the 1990-2030 period. Do these changes in dependency and support ratios tell us anything the growth of pension commitments, and of future pension liabilities? Figure 1 provides a simple scatter plot of the share of public pension payments in GDP in 1995 against the ADR in 1990. The polynomial drawn through the points indicates a weak positive relationship but there is a good deal of variability in the pension burden relative to the ADR that is unexplained by 9 At the risk of relying on anecdotal evidence, in a mniddle-income country visited by the author on a World Bank mission, published financial statistics contained figures of contribution revenues that were obtained by calculating notional employee and employer contribution liabilities. In practice, actual contribution receipts were, in one scheme, only 20% of these liabilities, and many state managed enterprises (SMEs) had not paid employee or employer contributions to the central exchequer for years. 11 demographic structure. Ageing does not seem to be the sole explanation of which countries have accumulated larger pension obligations through their public pension programmnes. It might be expected that a stronger (negative) relationship would be observed between future financial liabilities and projected changes in the weighted support ratio, since demographic trends are a key factor underpinning the simulated pension obligations. Figure 2 cross-plots the OECD projections of the absolute increase in net financial liabilities between 1995 and 2030 on the IBRD/World Bank estimnated changes in the needs-weighted support ratio from 1990 to 2030. As suggested, there appears to be a negative relationship. However, as the labelling illustrates, this is driven by three countries: Ireland with an improvement in the support ratio and no growth in pension liabilities and, on the other hand, Japan and, to a lesser extent, Finland, where a rapid deterioration in the support ratio is associated with a rapid rise in financial liabilities. For the remaining countries, large differences in the demographic projections are associated with a wide variation in the financial trajectory: ageing per se does not explain the cross-country variation in current pension progranmme commitments and the future deterioration of the fiscal positions.'" Figure 2. Change in aged dependency ratio and change in pension liabilities, 1995-2030 Ireand change in pension liabilites 2 (per cent of GDP) -5 -2 b 5 10 15 20 25 30 -4 chaxe in __ deperdery rato -14 -* (PerCeltage points) -16 - eands A more sophisticated approach might be to break down the changes into the support ratio into the impact of fertility (size of the cohort entering the labour market), longevity and labour force participation. 12 3.2 7he eay stages of WOfoPdsipon pvrlgrxVw For an augmented explanation of the detenroration of the fiscal status of public pension programmes up to the present time, we have to look at other factors. 'Te first explanation lies in the initial development of programmnes. Although many countries, especially in Europe and Latin America, developed public pension programmes in the inter- war period, the full introduction of comprehensive and universal social security took place after 1945. In the life cycle' of public pension programmes (as described in The World Bank, 1994, 315-317), the early years of schemes should be the years in which contributions are accumulated, assuming that pension entitlement is based on contributions rather than financed by transfers from the general budget 'Sound' actuarial practice involves the initial accumulation of a fund, to be dissolved at the maturation of the scheme, even in a scheme which in 'steady state' will be unfunded, in order to avoid rapid rises in contribution rates as the scheme reaches maturity. Given, too, that the first 'full' members of the contributory scheme were members of a large 'baby boom' generation, simple foresight would have strengthened this policy condusion to avoid the simultaneous consequences for contribution rates of both programmne maturation and population ageing. Although some public pension programmes did initially accumulate funds, and some of these remain, notably the US Social Security Trust Fund, foresight played little part in pension policy from the start. Initial trust funds either lost their real value through dubious investment decisions, or were invested in government securities, so acting as a cheap source of public credit, or were used to finance higher real pension commitments. Many govenmments decide to pay benefits out of the system to existing people reaching pensionable age even though such people had paid litde or nothing in to the programme. Such intergenerational redistribution could be justified on welfare grounds: the recipients of such generosity had, after all, lived through two world wars and a cataclysmic world recession. However, such policies destroyed any notional link between contributions and pension payments across generations, and thus any form of fiscal responsibility within the programmes. 13 3.3 Falling lawrfobn part nof older mnu A pervasive and well-known trend across OECD countries has been the decline in labour force participation of older men below normal state pension ages. (Ihis is illustrated in OECD, 1996, Charts 4.1 and 4.2)." Current employment rates for men aged 60-64 are under 20 per cent in France and the Netherlands, around a third in Italy and Germany, and around half in Australia, New Zealand, the United Kingdom and the United States. Japan has the most 60-64 year olds in jobs: around three-quarters. The implications of this for the support ratio have been concealed in many countries by a rise in the participation in paid work of married women. While participation after state pensionable age is declining largely as a result of pension programmes themselves, employer and governnent policies have encouraged the decline with disability progranmmes, relaxing job-search conditions for receipt of unemployment benefit, 'seniority pensions' (described previously) and various other 'special measures' (OECD, 1995). For example, a quarter of 60-64 year old men in the Netherlands and the United Kingdom receive disability benefits. 3.4 Forge onrs A somewhat surprising occurrence is that many governments have systematically underpredicted improvernents in longevity and overpredicted future fertility rates. It is not clear whether these errors arise as a result of incompetence or are politically motivated. For example, many projections of fertility rates tend to predict future stabilisation at close to replacement levels even when current rates are declining. There is also a reluctance to project longevity increases to continue at their current rate. For example, projecting the current decline in age-related mortality rates in the United States would suggest an average length of life of 85+ years by the middle of the 21't century. But the US Social Security Administration forecasts average life expectancy at only 81 years in the mniddle of the century, anticipating that age-related mortality improvements will slow considerably in the next few years. In contrast, other analysts have suggested improvements in life expectancy to accelerate, perhaps generating expected longevity of 100 years, over the same period (the evidence is summarised by Lee and Skinner, 1999). Such projections of rapid longevity improvement seem to be regarded as too 'radical' by official actuaries, who adopt rather For the United Kingdom, see Disney etal (1994) and Tanner (1998). 14 conservative assumptions. But of course, what is 'conservative' to the official actuary may prove wildly overoptirmistic from the point of view of the future fiscal sustainability of pension and health system if 'radical' projections turn out to be correct. In terms of consequences for pension reforn of misleading or omitted demographic projections, the past example of the United Kingdom is interesting, although it should be emphasised that I select this case merely on the grounds of proximity and particular knowledge. Unlike most other OECD countries, the United Kingdom has broadly maintained a scheme of flat benefits since 1946, supplemented by private pension schemes. In 1978, however, after a long political wrangle, the State Earnings Related Pension Scheme (SERPS) was introduced, to provide a comprehensive earnings-related pension for individuals who did not belong to private occupational schemes. Government Actuary projections of the additional costs of SERPS were only extended to the year 2005 when, with the 'baby-boom' generation yet to retire, the miore generous scheme looked affordable. Subsequent projections by Hemming and Kay (1982) and the Department of Social Security (1984) however took the forecast up to the year 2025, at which point it became apparent that the scheme would cost a large amount of additional money. For example, with earnings indexation of the basic state pension, the DSS projected the combined employee and employer contribution rate to rise from 15.9 per cent in 1985 to 23.2 per cent in 2025 as the scheme matured and the 'baby boom' generation retired. These findings played a large part in the subsequent policies of downgrading SERPS benefits and encouraging individuals to opt out of SERPS by buying personal pensions (Disney and Whitehouse, 1992). It is remarkable enough that a major pension programme was introduced without a proper assessment of the costs of the scheme either at its own maturity or at the maturation of the demographic transition. What is equally interesting is that, since that time, the official forecast of the number of retirement pensioners in the early part of the next century has steadily risen, even norrnalising for changes in the benefit regime since then. For example, in 1981, the number of retirement pensioners forecast for the year 2020 was 10.6 milion. By 1990, the forecast had risen to 13.4 million. By 1995 (ignoring the equalisation of pension ages for men and women during the decade 2010 to 2020, which is projected to reduce numbers by 2 million, in order to keep the calculations on a comparable basis), the figure 15 had risen to 14.4 million; an increase of 36 per cent.'2 Thus, had SERPS continued at its existing level of generosity, the required contribution rate at scheme maturity might now be projected to be as high as 31.5 per cent, not 23.2 per cent! For a country that had prided itself on avoiding the 'excess' commitments of other OECD countries, this episode was a salutary lesson. 3.5 7he Tonzz Game'natofur4 esxoiasewy The past and projected build up of the financial liabilities of the public pension programme has in large part depended on the political process underpinning the accrual of pension commitments by successive generations. To examine this factor in general terms, it is useful to start by considering the economic analysis that is normally applied to the problem of how to finance public pension schemes. A good deal of analysis of the relative merits of funded and unfunded social security has rested on the scheme satisfying the so-called 'Aaron-Samuelson' condition, named after seminal articles by Aaron (1966) and Samuelson (1958). Samuelson considered an economy where goods were perishable and where people sought to retire from producing their own consumption goods later in life. His point was that individual lifetime utility could be maximised if a 'social contract' could be arranged, so that each generation paid a 'pension' to each preceding generation, such that the implicit return on the contract was equal to the rate of population growth. Since the return on stonrng perishable goods was negative and population growth was likely to be positive, such an 'unfunded' scheme could be socially optimal. Aaron's generalisation of this 'rule' linked the equilibrium 'return' on unfunded social security to the rate of growth of earnings, being the sum of earnings growth per head and the growth of population. The mechanical application of the 'condition', in a world with capital, would simply compare the rate of return on capital (the return on a funded scheme) with the return on an unfunded scheme, as derived above. Where the latter was high, an unfunded scheme was superior to each generation simply relying on its own saving. Samuelson's paper bears careful re-reading, not least where he discusses the issue of how such a 'social contract' can be maintained. It is extremely hard to think of practical 12 These projections are taken from the successive Quinquennial Reviews of the Govemrnment Actuary's Department. The original costings for SERPS are contained in the Report by the Government Actuary on the 16 mechanisms by which such contracts can be replicated and implemented by successive generations, without imposing assumptions about the nature of transactions costs (Esteban and Sakovics, 1993), or the behaviour of agents (Hansson and Stuart, 1989). The key point of unfunded social security is that the financing of accruing liabilities is left to future generations. In that case, it bears much the same character as the schemes of Charles Ponzi, an originator of the use of chain letters to raise money."3 In particular, if the initial generation in an unfunded scheme obtains a 'retum' on its contributions in excess of the Aaron-Samuelson 'rule', later generations will have to accept lower returns to preserve the stability of the scheme. But it is tempting for subsequent generations to attempt to maintain the high initial rates of return by legislating over-generous benefit accruals for themselves, the liabilities for which will in turn be passed on to subsequent generations. The only constraint on this ratcheting up of programme liabilities is where a generation believes that a subsequent generation will simply renege on future commitments made by a prior generation. Ageing, by making such reneging behaviour more likely given the extra burden it imposes on workers, may actually make pension reform more likely.'4 There is a parallel in all this with the 'initial conditions' problem in the public choice literature. This suggests that 'excessive' government spending levels will be passed on from generation to generation, unless there are credible mechanisms to forestall the build-up of public comrnmitments. These mechanisms might include 'balanced budget' policy rules, tax ceilings or money supply controls. It is, however, hard to think of rules of this kind that can induce consistent behaviour in unfunded public pension programmes. Thus, differences im the build-up of pension obligations in unfunded schemes across countries largely reflect the presence or absence of mechanisms that enforce constraints on this short run optimising behaviour."5 financial provisions of the 1975 bill to introduce SERPS. 13 See Blanchard and Fischer (1989), note 23, p.84. 14 For an extended argument along these lines, and numerical illustrations, see Disney (1996), Chapter 9. i5 For example, election platfonms in some countries often include prormises to increase the generosity of benefit levels, lower the retirement age, and so on. These promises take litte account of the future extra liabilities incurred by such promises, which are rarely explicit in national accounts. A natural requirement to inform public debate would be to require that all reform proposals contained explicit accounts of their impact on future liabilities, and of tJe raticnal inidn of sd igs 17 4. Disparities in private provision Significant private provision of pensions, which is the main alternative vehicle of retirement income provision to public unfunded schemes, is not a universal rule among richer OECD countries. Examining the current and future potential importance of private pensions, there are two relevant questions. First, how large should private assets be to sustain similar levels of pension in retirement to those promised by unfunded programmes? Secondly, are funded programmes largely immunised from the demographic and political pressures facing unfunded public schemes? Table 3, drawn from OECD (1998a), provides details of private pension assets as a percentage of GDP for a number of countries. What fraction of GDP these assets should be to permit a 'substantial' private component to retirement income depends on demographics, the investment portfolio of private funds (which varies widely across countries) and the timing of maturation of funded schemes. Suppose we assume a 'support ratio' of 2, a desired replacement rate of pensions to lifetime average earnings of around 0.5, and a wage bill share in GDP of 0.7, then pension assets of around 140 per cent of GDP at scheme maturity would be required."6 Only three countries in Table 3 are even close to fulfiling this criterion: the Netherlands, Switzerland and the United Kingdom. It is no coincidence that all these countries' public programmes have largely flat-rate pension benefits. A further group of Anglo-Saxon countries (North America, Ireland and Australia) also have substantial private assets, followed by the Scandinavian countries. Few other countries among this group have private pension assets; New Zealand being the one country where legislative changes have eliminated a substantial private industry (for further details, see Stjohn, 1998 and Johnson, 1999). 16 This assumes that average earnings grow in real terms at 1% per annum, and that the average real return on the fund as it accumulates is 5% per annum. The pension is price indexed and we abstract from any tax treatment of income or assets. Out of interest, to generate a replacement rate of 0.5 on average lifetime earnings under these assumptions, individuals would need to save just under 9% of their earnings each year to achieve this target. 18 Table 3. Pension fund assets as a percentage of GDP, 1987 and 1996 Country 1987 1996 Switzerland 75 117 Netherlands 46 87 United Kingdom 62 75 United States 36 58 Ireland - 45 Canada 26 43 Japan 38 42 Finland 20 41 Average 27 33 Sweden 33 33 Australia - 31 Denmark 11 24 Luxembourg 20 20 Greece - 13 Portugal - 10 Norway 4 7 Germany 3 6 France - 6 Belgium 2 4 Spain - 4 Korea 3 3 Italy - 3 Austria - 1 Czech Republic - 1 Hungary - 0 Source: OECD (1998a), TableV.1 Note: Figures for Denmark include company pension funds only in 1996 and for Germany for 1993 and 1996 only. Figures for Finland cover financial assets only. First pillar assets are included in Sweden for 1987 and 1990. Average calculated over countries with both years' data only Are private and other non-social security programmes immune from the demographic transition? One problem is where supplementary pension commitments over and above the explicit social security programme are unfunded. In particular, pension schemes for public sector workers are often unfunded and, like other unfunded pension liabilities, not measured explicitly in the national accounts. This is largely true of the pension schemes of civil servants in the United Kingdom, France and Germany, for example (Bovenberg and Petersen, 1992). Efforts to measure the liabilities of public schemes often ignore these additional liabilities. Second, the future values of private pension funds are subject to some of the same pressures that are likely to affect public programmes. For example, the maturation of these 19 funds will accompany the demographic transition resulting from the baby boom generation reaching retirement age, and at that point funds will be net sellers, rather than buyers of financial assets. Whether this affects global asset values depends on the future path that other countries take in their pension arrangements and the extent to which portfolios are globally diversified. At present, a very small proportion of OECD pension assets are held outside the OECD area. This leaves open the possibility of large gains from further diversification of pension fund portfolios. Falling or stagnant asset values would make it that much more difficult for funded pension schemes to provide an adequate value of pensions. However, demographics are not fully synchronised across countries, and without further wholesale and immediate privatisation by all the OECD countries simultaneously, it is hard to believe that such global stagnation of capital market values will take place. A more pressing problem in a number of countries such as the United Kingdom, is a recent decline in ainuty rates, which may reflect uncertainty among insurers as to future life expectancy, imperfections in capital markets and the decline in long term interest rates. A more general issue is whether there is positive covariance across countries and time between the rate of return in capital markets (which determines the payoff of funded schemes) and the implicit potential rate of return on unfunded schemes (which is the rate of growth of the wage bill, or some similar measure). If, for example, falling labour-force growth is associated with falling returns in the capital market (perhaps because of a deepening capital stock), then funded schemes will not be immune from the problems of existing, unfunded, schemes. Studies which examine the rdame insurance properties of unfunded schemes versus funded schemes (such as, recently, Bohn, 1999) do not focus on this P.lio issue. Evidence on this latter issue is scanty. Palacios (1998) suggests that there is very little correlation between wage growth and equity returns over time. He examines the period from 1953-95 for Germany, Japan, the United Kingdom and the United States, and a somewhat shorter period for Sweden, and finds coefficients on a regression of annual wage growth on annual equity returns of between +0.047 and -0.01 for the different countries. None of these coefficients are significant. This suggests that diversification out of an unfunded scheme into some combination of funded and unfunded schemes should improve 20 the risk properties of the pension system. Indeed if other assets could be found that mimic the return structure implicit in linking part of the pension scheme to wage growth, a funded component (subject to any redistributive aims) might seem unnecessary. These results are interesting, but cover rather short time periods and a sub-set of countries: a full analysis would need to look at the covariance across countries as well as between countries, and examine portfolio holdings. Nevertheless, the analysis in this section does nothing to refute the proposition that some shift towards private provision might guarantee financial security relative to the current over-reliance on unfunded pension programmes. 5. A taxonomy of pension reforms It is now useful to consider what is 'on the table' in terms of pension reform across OECD countries. Rather than proceed on a country-by-country basis (see, for example, Johnson, 1999, for such an approach), it is possible to construct a basic typology of such reforn proposals. A key distinction is obviously between proposals that maintain a strong unfunded element, and those that emphasise the virtues of a transition to a funded programme. However the 'funded versus unfunded' debate may have become overwrought. Mixed schemes which contain funded and unfunded elements, such as the pension programme in the United Kingdom, may of course offer a plausible 'half way house'. Indeed, it is hard to find a funded proposal that does not contain, at the very least, a residual, publicly financed safety net. Similarly even the harshest critic of fully funded programmes generally accepts that individuals should have the right to engage in supplementary saving for retirement in financial assets if they so choose. In similar vein, a mixed tier or 'multipillar' approach is central to the pension strategy advocated by the World Bank (1994). However, the risk of mixed strategies is that the pension programme becomes excessively complicated. The United Kingdom, in which individuals can choose in their second tier mandatory provision between a public pension benefit (the state earnings-related pension, Serps), a company-provided occupational pension, or an individual retirement saving account (a 'personal pension), is a case in point. Such choice-based systems require a high degree of transparency and individual knowledge of pension accrual structures (net of transactions costs) if the risk of people choosing unwisely is to be avoided, as the personal 21 pension 'mis-selling' scandal has illustrated.'7 The recent pension reform proposals (Department of Social Security, 1998) would complicate the choice still further by introducing a fourth route: the 'stakeholder pension'. The complexity engendered by mixed strategies is illustrated by the sylised chart (Figure 3) of what pension provision may look like in the imrnediate future in the United Kingdom if the current proposals are implemented (see Disney, Emmerson and Tanner, 1999 for further details). Figure 3. The pension system in the United Kingdom following the 1998 reform Third tier: voluntary Additional voluntary Free-standing Pooled pension Other savings: contributions (AVCs) AVCs investment (PPI or Lisa) ISA, life insurance, OEICs, unit trusts etc. ................. .. ............................... . .. ........ ..................................... ......... ..................................... .................. Second tier: mandatory, choice of options _ F, :I , IF =,s Occupational pension Personal pension Stakeholder pension State second pension (DB and DC) (individual or group) ..................................... ............ .. ......................................................................................................... Basic state Minimum income pension (flat rate) guarantee (means-tested) First tier: mandatory, public In what follows, four generic reform strategies are considered. The first two involve retaining a strong unfunded component. These are denoted a 'parametric' reform strategy (after Chand and Jaeger, 1996) and a strategy based on 'actuarially fair' public pension programmes respectively. The second two strategies involve a strong funded private component and are denoted as a 'clean break' privatisation and a 'partial' privatisation. In the latter case, where not all individuals join the privatised programme, a key issue is whether it is the government or individuals themselves that decide who can join the private funded scheme. 17 See Whitehouse (1998, 1999) for a discussion of the mis-selling issue. 22 5.1 A paramtric'mrfom of the u*xWprcgran Chand and Jaeger denote a reform of this type as 'parametric' because, presumably, the key choice variables of an unfunded scheme can indeed be written down as a function of a few 'parameters'. In the standard approach to financing an unfunded pension programme, define B = number of beneficiaries, L = number of workers, w = average wage rate, p = average pension and c = contribution rate. Then, pay-as-you-go equilibrium requires that c = (B/L)(/w) where (B/L) is the inverse of the support ratio and (plw) is the average replacement rate. Clearly to reduce c (or to avoid growing debt if c is fixed by, say, law at below its equilibrium level) requires any of, or a combination of, reducing p or B, or raising w or L. Increasing the support ratio requires either reducing B or increasing L. Raising the state pensionable age and reducing entitlements for dependants are ways of reducing B. Raising L requires an increase in the economically active proportion of the population with policies such as lower early retirement pensions and stricter regulation of other 'routes' to inactivity (e-g. disability benefits). Reducing the replacement rate means either reducing benefits directly or indirectly by reducing the generosity of post-retirement indexation procedures, or raising eligible wages by for example increasing the fraction of the wage, or the wage bill, that is liable to pay contributions. Various simulations of 'parametric' reforms of this type, for a number of OECD countries, have been carried out by international organisations, such as in Chand and Jaeger (1996), with a view to examining their impact on fiscal liabilities. They are (or should also be) standard in projections carried out by official actuaries when estimating the sensitivity of the future costs of pension programmes to changes in baseline assumptions. Most such simulations find that the key parameter in determining fiscal liabilities is the state pensionable age (often equated with the 'retirement age' in such calculations). Raising state pensionable age, or perhaps more specifically linking it explicitly to expected longevity, is 23 generally a key policy in 'parametric' reforms to the problem of financing public pension programmes. A number of OECD countries are changing pension ages.'8 Although these policies are uncontroversial from an economic point of view (although rarely so from a political perspective), the problem lies in modelling gains from them. A key objection to analyses of the gains, in terms of the projected reductions in fiscal liabilities, is that retirement behaviour is rarely modelled explicitly in such simulations (as indeed in the equating of 'retirement' with the first age of receipt of benefits from the public pension progranmme). While raising state pensionable age does indeed reduce the denominator of the support ratio directly, the indirect impact on the number of beneficiaries is far from clear. Employment (and even participation) rates in the years prior to reaching state pensionable age are well below 100 per cent in most OECD countries, and simply raising state pensionable age still further may not necessarily increase participation rates, let alone employment rates, among older workers. If these extra people below pension age thereby end up on unemployment or disability benefits, or other forms of welfare support, these costs will offset in budgetary terms the 'gain' in the reduction in the number of explicit beneficiaries of the public pension programme. Estimates of the benefits of 'parametric' reforms which do not model these behavioural effects directly are systematically biased in favour of finding large reductions in fiscal liabilities from such reforms precisely because they treat as parameters what are, in fact, behavioural variables. 5.2 An 'a fair`w'oai p xrne A second broad strategy for reform of an unfunded programme, which leaves the financing strategy intact, is to link benefits and contributions explicitly, for each generation, to the Aaron-Samuelson sustainable 'return' to an unfunded scheme. Such a strategy, in general terms, calculates what is the sustainable implicit 'rate of return' on the contributions of each cohort of contributors. This depends on the projected growth or decline in the real contribution base (usually proxied by the wage bill). The accrual rate on pensions is then set so that this return is, on average, realised. This policy lies behind the so-called 'Dini reform' 18 In the US, the age of full entitlement is to be raised to 67 years by 2020. Finland Italy, New Zealand and Spain are also phasing-in increases in pensionable age. Other countries, such as Australia, Germany and the United Kingdom are raising women's pension age to equalise it with men's. See Disney and Whitehouse (1999a), Table 5 and OECD (1998a), Figure IV.1. 24 of the Italian programme in the rnid-1990s, and behind the public assertion that such a reform puts the programme on an 'actuarially fair' basis. Since 'actuarial fairness' also involves questions of redistribution witbr generations, or cohorts, such a reform has to include subsidiary but important modifications of plans: towards, for example, an average lifetime basis for calculating pension entitlements. The move from a 'final' or best salary basis will obviously induce some variation in outcomes if earnings profiles are subject to volatility.'9 At its logical' extreme, in the so-called 'notional accounts' variant of this strategy, each individual pension is supposed to be explicitly based on contributions such as to minimise inter-individual variation in returns. In the recent Swedish reform and in similar reforms elsewhere, such as Latvia and Poland, each individual is given a 'notional account' within the public pension programme. This mimics a funded retirement saving account, with the important difference, of course, that the accounts are 'notional' (i.e. not funded). With a linear accrual structure, the incremental accrual of pension benefits should be transparent.20 Implicit in a reform of this type is that the failures of unfunded programmes up to now arise because of their lack of transparency and from their inability to apply a rigid formula linking returns to the Aaron-Samuelson condition. These policies would have minimised arbitrary redistribution arising from the vagaries of the benefit formula and political processes, it is argued. Since this reform strategy is currently rather fashionable, a few difficulties need to be pointed out. First, there are still no plausible enforcement mechanisms to guarantee that the 'Aaron-Samuelson' condition for equilibrium in an unfunded programme will be satisfied in the future. In the Italian reform, much of the generational burden of shifting to this formula-based approach falls on younger generations. There has been a conspicuous lack of success in the Italian reform in cutting back pension benefits in the immediate future, whereby the burden would fall on prime age and older generations. In the Swedish or Latvian-style reform, ingenious adjustment of the pension indexation formula will be used to keep the fiscal outcome 'on track'. Suppose, however, that a temporary recession reduces real wage bill growth becomes zero. Irrespective of, say, 19 See the discussion in Disney and Whitehouse (1999a), section 8.2. 20 See Fox and Palmer (1999) on Latvia, Gora and Rutkowski (1998) on Poland, Palmer (1999) and Sunden (1999). 25 inflation, nominal pensions will be held constant. Politically, such an outcome seems unlikely: even the United Kingdom, with its low level of public pension provision, automatically indexes benefits to inflation. There is likely to be a ratchet effect by which, when the wage bill rises faster than trend, pension are raised more rapidly but, where wage bill growth falls below 'headline' indicators such as the rate of inflation, pensions are raised in line with inflation. As a result, fiscal liabilities will continue to grow relatively rapidly. Finally, by attempting to elirninate much of the redistribution inherent in unfunded schemes, so as to make the benefit-contribution link 'transparent', much of the rationale for the public programme is destroyed. Public programmes inherently redistribute from poor to rich (because the poor die younger), from men to women, from married couples to widows, and from the rest to the very poorest through benefit 'floors'. Moreover, participants can automatically compare explicit 'returns' to such accounts with the much higher returns to be obtained in marketed private savings accounts. Will individuals be willing to contribute to a programme which may well explicitly offer a negative return on contributions to later generations2' while observing possible double digit nominal returns on private saving accounts? It is not overly cynical to suggest that it is the lack of transparency of unfunded public pension schemes, coupled with their undoubted insurance-based and redistributive components, that encourages people to contribute to such programmes. Greater 'transparency' only makes their inadequacies more transparent, and the programmes less attractive. 5.3 ICZ=-bLak'priin The alternative strategy to 'fixing up' the unfunded pension progranmme is to replace much of it with a funded programme. Such an approach has not been implemented explicitly in OECD countries. The nearest variants are the introduction of mandatory superannuation in Australia in 1992, and the 1986 reform in the United Kingdom, which downgraded SERPS and allowed individuals to 'opt out' of that scheme into an individual retirement savings account known as a 'personal pension'. The latter reform is however more appropriately considered under the next sub-heading. 21 See, for example, Hurd and Shoven (1985) for an empirical study of different cohort's returns from social security in the United States and Disney and Whitehouse (1993) on the United Kingdom. 26 While in principle the government can introduce or extend its own managed funded component to the public programme, a generic transition to a funded programme is almost certain to include a private component, and thus I refer to this as 'privatisation'. The big objection to allowing the government to organise the funded programme itself is that it is extremely difficult to 'ring fence' the public pension fund from other components of the government budget.22 Suppose, for example, as in the United States, that the assets of the public pension programme's 'trust fund' are held in the form of governnent securities. Insofar as the purchase of government securities by the fund permits the government to increase its own liabilities by, for example, increasing its spending uncovered by tax receipts, then the growth of government liabilities matches the growth of government assets held by the fund. The funded and unfunded programmnes are then equivalent in their generational incidence (Kotlikoff, 1992).23 For this reason I consider only a funded option which is combimed with privatisation. The most well known full scale privatisation of a public pension programme is that which took place in Chile in November 1980 (see Edwards and Edwards, 1991; Edwards, 1998), which has been copied in a number of Latin American countries and elsewhere.24 Some influential commentators are advocating reforms of this kind in OECD countries (B6rsch-Supan, 1998a; Feldstein, 1996, 1998). The attractions of full funding of the programme are as follows. First, the real return on funding almost always exceeds the 'Aaron-Samuelson' return on unfunded schemes, even if one can argue over the appropriate private rate of return to use.25 Second, a funded scheme is transparent, not in the artificial sense of 'notional accounts', but in the real sense that benefits are explicitly related to contributions and capital market performance rather than to some formula of the public programme. In contrast, contributions to an unfunded 22 See Iglesisas and Palacios (1999) for a detailed survey of the performance of publicly mandaged pension assets. 23 This is of course a whole sub-debate in the literature. What would happen, for example, if we let the government pension fund hold equities and other private sector assets? This would imply that other private asset holders would own a greater fraction of government securities while the governnent would hold a greater share of the equity market. The effect of this on portfolio values depends on what we assume about financial markets. There is also the rather obvious question of whether the governrnent is a better fund manager than the private sector. Again, see Iglesias and Palacios (1999) on this last issue. 24 See, iralia, Queisser (1998) for a survey of Latin American reforms. 25 For example, the rate should account for transactions costs in establishing funded accounts, net out the risk premnium and adjust for tax liabilities. 27 public pension programme inevitably contain a tax component, which distorts labour supply and savings behaviour relative to saving in a funded programme. Note that this tax distortion has two dimensions. First, individuals are forced to buy a given level of longevity insurance (pensions) at an excessive cost given the low return to the unfunded scheme. Secondly, the pattern of individual benefits implies different rates of return on individuals' contributions and thus there are arbitrary transfers of income between contributors. Feldstein (1998, 'Introduction') has estimated the former loss at 1 per cent of GDP in the United States, which is very large relative to most 'welfare triangles' calculated in similar exercises. In a 'clean break' privatisation of the pension programme, no further contributions are made into the existing unfunded programme. All new contributions after privatisation are made to the pension funds and are assigned to individual pension accounts. However, the liabilities of the unfunded programme, which comprise both existing and projected payments to current pensioners and the accrued pension rights of those who have not yet retired within the unfunded programme, have then to be financed by some means. This is the crux of the funding transition problem: the implicit liabilities of the existing programme then become explicit and are supplemented by the additional liabilities arising from the transfer of all future contributions to the new, funded, scheme. Of course, the assets of the new funded programmes may match, or be less than or more than, the additional liabilities arising from the future cessation of contributions to the unfunded programme. As in Chile, one of the main purchasers of the extra liabilities of the governnent may be the new pension funds. This offsets, at least partially, the 'extra' financing liability. Nevertheless, the immediate impact of privatisation is an explicit jump in the liabilities of the public sector and, to those that do not like large public sector liabilities, such as the IMF, this issue seems to be the insurmountable hurdle of privatisation.26 There are, however, several ways of handling the transition issue. One is simply to accept the extra explicit burden of public liabilities and a higher perpetual burden of interest payments on the debt. Chand and Jaeger (1996), Table 18, estimate the increase in pension liabilities from this strategy in a range of industrial countries to be on average 152 per cent of 26 Chand andJaeger (1996). See also Holzrmann (1998ab) on the transition issue. 28 1995 GDP. Sixty per cent of this increase would arise simply from making explicit existing unfunded liabilities that are not currently covered by legislated contribution rate increases. It would be possible, in principle, to model (heroically) the implications for real interest rates under various assumptions. The second alternative is to have an explicit transition finance strategy that effectively involves establishing a generational incidence of the transition burden. For example, Kotlikoff (in Feldstein, 1998) uses the Auerbach-Kotlikoff (1987) GE model to examine a transition to full funding in the US economy. He adopts three transition- financing scenarios: lump sum transfers to compensate losing generations, an increase in income tax rates or a new consumption (expenditure) tax. The last gives the strongest aggregate welfare gain in the model sirnulations. Another transition strategy, where the generational incidence is less clear, is where the liabilities are in part financed by higher budget surpluses over a substantial period. These are achieved by, say, cutting public capital spending or by privatisation of public assets. In the long run, the economy will have a higher stock of private assets in the pension funds and a lower stock of public assets; the welfare consequences of such a transition are more difficult to evaluate without knowledge of the incidence of the benefits from publicly owned assets. Another common objection to a funded solution is that it rules out any explicit redistribution as part of the public pension programme. While elimination of arbitrary redistribution may be desirable between generations, it reduces the equity component of such programmes uitin generations. A particular problem is that, in the face of transactions costs, persistent low-income earners may be able only to earn very low pensions with their contributions. An obvious solution is for the government to provide a safety net or some explicit minimum pension guarantee within the funded programme. However, the costs of the safety net have to be added to the fiscal liability of the programme, and generous guarantees may encourage low-income contributors to opt out (either formally, if permitted, or informally, if the scheme is compulsory). Such non-participants may reason that the returns on their contributions, less the deduction of the public guaranteed payment that will obtain if they can raise their pension by saving, make continued contribution to the programme unprofitable. This is a trade-off that has to be evaluated explicitly in the privatisation route. 29 All OECD countries with a strong funded component to pension provision do in fact supplement private provision with a flat or income-tested unfunded component. Chile and other Latin American reforners also have either a minimum pension guarantee or social-assistance top-ups for people with low incomes. This combination of funded and unfunded components tends to emphasise the redistributive impact of the public component, without adversely affecting overall pensioner incomes. This finding is illustrated from research using rnicro data undertaken as part of the OECD's programmne of work on ageing. As Figure 4 shows (see also OECD, 1998, Table IV.3), ratios of average pu c pesion income to pre-retirement income vary widely across OECD countries but ratios of tota pesion innre to pre-retirement income are very similar. In contrast, in the lowest quintile of the income distribution, as the second panel of Figure 4 shows, the contribution of public pension programmes to retirement incomes is very similar across countries." 27 Figure 3 gives the strong implication that comprehensive public provision has crowded out other, private, forms of retirement income. See B6rsch-Supan (1998) for further discussion. It does not of course follow that reductions in public pension benefits would automatically generate a private pension mnarket. 30 Figure 4. Ratio of pensioners' incomes to workers' incomes overall average total disposable income Germany K . public transfers France Netherlands Sweden United Kingdom 0 20 40 60 80 100 replacement rate, per cent lowest income quintile total disposable income United Kingdom . -ublic transfers France -. , Germany _ -;-&-- Netherlands Sweden1 x _ - _ . _.: 0 10 20 30 40 50 60 70 80 90 100 replacement rate, per cent Note: Married couples. Head of 'pensioner households aged 67 and 'working-age' households aged 55 Source: Disney, Mira d'Ercole and Scherer (1998) A third argument often presented against funded schemes of this kind is that they subject participants to investment risk. I would rather restate this as suggesting that funding replaces potential political risk with potential investment risk. The impact of political risk should not be understated. Table 4 shows how pension benefits have been affected by recent reforms in six major OECD countries. Benefits are measured by 'social security 31 wealth', the net present value of the stream of pension payments. The calculations, on a common basis across countries, are for a 45 year old man on average earnings.28 In both relative and absolute terms, the largest cut in benefits is from Italy's 'Amato' reforms of 1992. Before the reform, social security wealth was estimated at 8.4 times average earnings. After the reform, this ratio was estimated at 51/4 times, a cut of over three times earnings or 38 per cent. The United Kingdom's two reforms of 1986 and 1994 cumulatively cut benefits by 27 per cent, a similar impact to the 1983 reform in the United States. Changes in the other three countnres have been rather smaller. Post-reform social security benefits are around 180 per cent of average earnings in the United Kingdom and 125 per cent in the United States. These are substantially lower than other countries, as would be expected from the data in Tables 1 and 2: over 400 in France and over 300 in Germany and Japan. Table 4. Impact of pension reforms on social security wealth Absolute change in Percentage change in social securty wealth social security wealth (per cent of eamings) France - assessment period 10 to 25 years -74 -14 (1993) - post-retirement indexation wages to prices Germany - pension age 63 to 65 -26 -7 (1992) - post-retirement indexation gross to net wages Italy - pension age from 60 to 65 -316 -38 (1992) - assessment period 5 to 10 years - pre-retirement indexation wages to prices+1 % - post-retirement indexation wages to prices Japan - first tier pension age 60 to 65 -66 -15 (1994) - post-retirement indexation wages to prices for second ter United Kingdom - assessment period 20 years to working life -52 -23 (1986) - replacement rate 25 to 20% United Kingdom - higher flat-rate deducton from eamings for -9 -5 (1994) second ter pension United States - pension age 65 to 67 -40 -25 (1983) - some taxabon of benefits - one-off indexation pause Note: Changes in pension ages and social-security wealth data are for men. Changes for women are in almost every case larger because of their longer life expectancy. Also women's pension ages have risen by more than men's in Germany and will increase to match men's in the United Kingdom Source: McHale (1999), Tables 5 and 6a 28 In fact, the earnings of the average production worker. See OECD (1998b). 32 Evaluating the relative importance of political, investment and other kinds of risk is problematic. It can be argued, for example, that an explicit funded strategy allows the individual to dic5 risk: the observation that equity markets are 'risky' is an irrelevance to the debate. There are two key questions. First, whether, by adding a portfolio of market assets to their pension 'portfolio', individuals can thereby reduce the prospective 'riskiness' attached to their future pensions. Secondly, whether the market can provide types of insurance against annuity risk arising from the volatility of markets. Simply asserting that a funded pension is 'risky' because it may be (in part) invested in equities is erroneous reasoning. There are, however, more plausible difficulties associated with investment risk, which need, at the very least, low cost technical solutions. One is the process of annuitisation itself, which will imply a change in the portfolio holding of the individual as the accumulated fund is converted into an annuity stream. If the point of conversion is simultaneous with an adverse outcome arising from market volatility, the annuity stream will be of lower value. However, deferral of annuitisation in such circumstances raises a real problem of adverse selection. In general, the annuity market itself, which may have been rather thin prior to the decision to privatise a component of the pension programme, may have more general adverse selection problems and may need careful regulation. A large compulsory arnnuity market, however, might be expected to reduce adverse selection problems relative to a voluntary market.29 Finally, annuity rates will continue to fall as longevity increases, and especially if there is greater uncertainty as to longevity given the potential impact of medical improvements and improved nursing care (see the discussion in section 3.4 above). 5.4 Partalpiwisatioz A proposal that has gained attention more recently is to privatise the programme partially. 'Partial' in this context means not just the possibility of maintaining an unfunded component to the programme as a whole: other possibilities include only allowing certain individuals to join the funded scheme, or alternatively of allowing individuals the choice of joining a funded or unfunded programme. The latter, choice-based option is illustrated 29 Walliser (1997, 1998) explores this issue from a theoretical perspective. Finelstein and Poterba (1999) examine the annuities market in the United Kingdom, and find that the losses from adverse selection in the compulosry part of the market are around half of the adverse-selection cost with voluntary annuities. 33 among OECD countries by the United Kingdom, in which individuals can choose to remain in SERPS (the unfunded component of the programme) or to invest part of their national insurance contribution in a personal pension." An example of the former, cohort-based option, is illustrated by 'basic pension plus'. This scheme, proposed by the Conservative administration ae' the 1997 general election, would have replaced the basic state pension with a funded programme for new cohorts entering the labour market. Other countries that have used partial privatisations, which either include a voluntary element, or a cohort- specific transition, include Argentina, Hungary and Poland.32 The essential point of such a reformn is that it reduces or spreads the transition costs more widely. If onhl younger workers enter the funded programme (whether through compulsion or choice), then they can obtain reasonable pensions on retirement by the sirnple effect of compound interest on their accumulating fund. At the same time, if earnings rise with age, the initial transition cost of their foregone contributions to the public unfunded programme is low. Over time, the value of foregone contributions rises but, of course, the fiscal liability will also peak. In contrast, in a 'clean break' privatisation, the cost of attempting to replicate the generosity of the 'defined benefit' programme for all participants may involve an immediate large increase in financial liabilities, as has been shown previously. A different problem emerges with partial privatisation if there are large within- generation variations in funded returns, for example due to transactions costs, or differences in individual entitlements in the unfunded programme. Suppose that participation is voluntary. It is possible, for example, that al rich people within a cohort might opt to join the private programme while all poorer people Qpted to stay in the unfunded scheme. This is more likely if the unfunded programme contained various explicit redistributive elements such as benefit ceilings or non-linear benefit accrual rates with income (the extreme example of which is a flat component to benefits). This generates a standard adverse selection problem, and in addition the existence of the funded option limits the capacity of the 30 Kotlikoff, Smetters and Waliser (1998) and Gusnan and Steineier (1995) look at the effects of mandatory and voluntary participation in private pensions as a substitute for social security in the United States. 31 See Whitehouse (1998, seaion Vi), Whitehouse and Wolf (1997) and Department of Social Security (1997) for a description and analysis on this proposal. 34 unfunded scheme to redistribute within generations. On the other hand, allowing only some individuals to join the funded scheme (differentiated, for example, by age) may lead to conflict, especially if returns to the funded scheme are well in excess of those to the unfunded scheme, making the funded option attractive to a much wider group of people. There is also the simple point in voluntary privatisations, such as the United Kingdom, that many people misunderstand accrual structures and therefore make 'incorrect' choices. 6. Conclusion The preceding discussion has suggested that public pension programmes have generated excessive financial liabilities for a number of reasons, of which the ageing of the population is only one among several reasons. There are a number of solutions on offer but, not surprisingly, they all have weaknesses. There are some obvious reforms that can be carried out within existing unfunded schemes, although a switch to a scheme of 'notional accounts' would seem to be an unnecessary complication and, ultimately, a dead end. The long run solution seems to involve a strong funded element, but the transition costs have to be handled carefully and partial strategies need to take particular care as to their distributional cornsequences and potential for complexity. 32 These issues are explored in Disney, Palacios and Whitehouse (1999) and Palacios and Whitehouse (1998). 35 Table Al. Estimates of dependency ratios, 1960-2030 Country Elderly dependency ratio Total dependency ratio Needs-weighted support ratio 1960 A60-90 1990 A90-30 2030 1960 z60-90 1990 A90-30 2030 1960 A60-90 1990 490-30 2030 United States 15.4 +3.7 19.1 +17.7 36.8 67.4 -15.7 51.7 +16.3 68.0 63.7 +4.0 67.7 -8.5 59.1 Japan 9.5 +7.6 17.1 +27.4 44.5 56.6 -13.1 43.5 +30.0 70.5 68.5 +2.6 71.1 -14.1 57.0 Germany 16.0 +5.7 21.7 +27.5 49.2 47.4 -2.1 45.3 +29.8 75.1 70.0 -0.8 69.2 -14.0 55.2 France 18.8 +2.0 20.8 +18.3 39.1 61.3 -10.2 51.1 +18.8 67.9 64.7 +2.8 67.5 -8.8 58.7 Italy 13.3 +8.3 21.6 +26.7 48.3 47.9 -2.4 45.5 +27.2 72.7 70.5 -1.4 69.1 -13.2 55.9 United Kingdom 17,9 +6.1 24.0 +14.7 38.7 53.7 -0.8 52.9 +15.1 68.0 67.3 -1.2 66.1 -7.4 58.7 Canada 13.0 +3.7 16.7 +22.4 39.1 70.5 -23.0 47.5 +21.5 69.0 63.3 +6.6 69.7 -11.3 58.4 Australia 13.9 +2.1 16.0 +17.0 33.0 63.2 -14.3 48.9 +13.7 62.6 65.3 +4.1 69.4 -8.2 61.2 Austria 18.6 +3.8 22.4 +21.6 44.0 52.1 -3.9 48.2 +23.2 71.4 67.7 +0.3 68.0 -9.0 57.0 Belgium 18.5 +3.9 22.4 +18.7 41.1 55.0 -5.8 49.2 +19.7 68.9 66.7 +1.0 67.7 -9.6 58.1 Denmark 16.5 +6.2 22.7 +15.0 37.7 55.8 -7.9 47.9 +19.1 67.0 67.0 +10 68.0 -8.8 59.2 Finland 11.7 +8.0 19.7 +21.4 41.1 60.6 -12.2 48.4 +22.5 70.9 66.6 +2.0 68.6 -11.0 57.6 Greece 12.3 +8.9 21.2 +19.7 40.9 52.0 -2.4 49.6 +16.9 66.3 69.3 -1.4 67.9 -9.2 58.7 Iceland 14.1 +2.5 16.6 +15.5 32.1 75.0 -19.8 55.2 +8.0 63.2 61.8 +5.4 67.2 -5.9 61.3 Ireland 18.6 -0.2 18.4 +6.9 25.3 70.6 -9.2 61.4 -6.9 54.5 62.1 +2.7 64.8 +0.5 65.3 Luxembourg 15.9 +4.0 19.9 +24.3 44.2 47.4 -2.6 44.8 +28.9 72.7 70.0 -0.2 69.8 -13.2 56.6 Netherlands 14.7 +4.4 19.1 +26.0 45.1 63.9 -19.4 44.5 +28.7 73.2 64.9 +5.4 70.2 -13.9 56.3 New Zealand .. .. 16.7 +13.8 30.5 .. .. 50.9 +10.7 61.6 .. .. 68.6 -6.5 62.1 Norway 17.3 +7.9 25.2 +13.5 38.7 58.2 -3.8 54.4 +13.9 68.3 66.1 -0.8 65.3 -6.6 58.7 Portugal 12.7 +6.8 19.5 +14.0 33.5 59.1 -8.4 50.7 +9.1 59.8 66.9 +1.0 67.9 -6.0 61.9 Spain 12.7 +7.1 19.8 +21.1 41.0 55.1 -5.8 49.3 +15.5 64.8 68.2 +0.1 68.3 -9.2 59.1 Sweden 17.8 +9.8 27.6 +11.8 39.4 51.8 -3.5 55.3 +15.1 70.4 68.0 -3.5 64.5 -6.5 58.0 Switzerland 15.5 +6.5 22.0 +26.6 48.6 51.5 -5.6 46.1 +309 77.0 68.7 +0.1 68.8 -13.9 54.9 Source: Bos et a/. (1994) 36 8. Bibliography Aaron, Hj. (1966), 'The social insurance paradox', Cadimn Jounal of Econcmrws, vol. 32, pp. 371-374. Auerbach, AJ. and Kotlikoff, LJ. (1987), Dynami Fiscal Policy, Cambridge University Press. Blake, D. (1995), Pension Scms and Pension Funds in the UnitedKktnf , Clarendon Press, Oxford. Blanchard, OJ. and Fischer, S. (1989), Lectures on Maoaxwanics, MIT Press, Cambridge, Mass. Bohn, H. (1999), 'Social security and demographic uncertainty: the risk sharing properties of alternative policies', National Bureau of Economic Research Working Paper no. 7030, Cambridge, Mass. Bbrsch-Supan, A. (1998a), 'Germany: a social security system on the verge of collapse', in Siebert, H. (ed.), Radesigaing SoIa Seami, Mohr Siebeck, Tiibingen. - (1998b), 'Retirement income level, risk and substitution among income components', Ageing Working Paper no. 3.7, OECD, Paris. Bos, E., Vu, M.T., Massiah, E. and Bulatao, K (1994), WAd Populti Prjecions, 1994- 94, World Bank, Washington, D.C. Bovenberg, L. and Petersen, C. (1992), 'ublic debt and pension policy', Fiscal Studies, vol. 13, no. 3, pp. 1-14. Chand, S.K and Jaeger, A. (1996), 'Aging populations and public pension schemes,' Occasional Paper no. 147, Intemational Monetary Fund, Washington, D.C. Chlon, A., Gora, M. and Rutkowski, M. (1999), The quest for pension reform: Poland's Secuity hgb Diwrsity', Pension Reform Primer series, Social Protection Discussion paper no. 9815, World Bank, Washington, D.C. Cutler, D.M, Poterba, J.M., Sheirier, L.M. and Summers, L.H. (1990), 'An ageing society: opportunity or challenge?' Brookirg Papers on EcoanicActvity, vol. 1, pp. 1-73. Demarco, G., Rofman, R., with Whitehouse, E.R. (1998), 'Supervising mandatory funded pension systems: issue and challenges', Pension Reform Primer series, Social Protection Discussion Paper no. 9817, World Bank, Washington, D.C. Disney, RF. (1996), GC We Afford to Grow Older? A Perspeaim on the Ecoarnxs of Agig, MIT Press, Cambridge, Mass. - (1999), 'Notional accounts as a pension reform strategy: an evaluation', Social Protection Discussion Paper, Pension Reform Primer series, World Bank, Washington, D.C., forthcomning. -, Emmerson, C. and Tanner, S. (1999), Panershi n Pess: An Assesn, Commentary no. 78, Institute for Fiscal Studies, London. -,ira d'Ercole, M. and Scherer, P. (1998), 'Resources on retirement', Ageing Working Paper no. 4.3, OECD, Paris. -, Meghir, C.HD., and Whitehouse, E.R. (1994), 'Retirement behaviour in Britain', Fiscal Stdies, vol. 15, no. 1, pp. 24-43. 37 Palacios, Rj. and Whitehouse, E.R. (1999), 'Individual choice of pension arrangement as a pension reform strategy', Working Paper no. 99/18, Institute for Fiscal Studies, London (also, forthcoming in the Scdvian Joumal of Ecoreins). -and Whitehouse, E.R (1992), 7hePersnalPevsion Stamnpde, Institute for Fiscal Studies, London. -and - (1993), Will younger cohorts obtain a worse deal from the UK state pension scheme?' in Creedy, J. and Casson, M. (eds), Indutnal C cmtration and Econanc Ineqlity, Aldershot: Edward Elgar. - and - (1999), 'Pension plans and retirement incentives', Pension Reform Primer series, Social Protection Discussion Paper, World Bank, Washington, D.C., forthcoming. Edwards, S. (1998), The Chilean pension reform: A pioneering program' in Feldstein, M. (ed.), Prtizing Sc&al Searrity, Chicago University Press. - and Edwards, A. (1991), Moneaisn and Liberalizatinbr The CGlaw Exp&inmt 2nd Edition, Chicago University Press. Esteban, J.M. and Sakovics, J. (1993), Intertemporal transfer institutions', Ja4vnl of Econami Thw7y, vol. 61, pp. 189-205. Feldstein, M. (1996), The mnissing piece in policy analysis: social security reform', Amemika ERcoanicRRiew Papers and P , vol. 86, May, pp. 1-14. - (1998) (ed.), PAizg Socil Seam5y, Chicago Uriversity Press. Finkelstein, A. and Poterba, J. (1999), 'Selection effects in the market for individual annuities: new evidence from the United Kingdom', mnw., Massachusetts Institute of Technology. Fox, L. and Palmer, E. (1999), 'Latvian pension reform', Social Protection Discussion Paper, World Bank, Washington, D.C., forthcomning. Geanakopolos, J., Mitchell, O.S. and Zeldes, S.P. (1998), 'Social security mone,'s worth', Working Paper no. 6722, National Bureau of Economic Research, Cambridge, Mass. Grandolini, G. and Cerda, L. (1998), 'The 1997 Mexican pension reform', Policy Research Working Paper no. 1933, World Bank, Washington, D.C. Gustman, A. and Steinmeier, T. (1995), 'Privatising social security: first round effects of a generic voluntary reform', Working Paper no. 5362, National Bureau of Economic Research, Cambridge, Mass. Hansson, I. And Stuart, C. (1989), 'Social security as trade among living generations', Ameria EonacRe, vol. 79, December, pp. 1182-1195. Heller, P.S., Hemming, R. and Kohnert, P.W. (1986), 'Ageing and social expenditures in major industrialised countries', Occasional Paper no. 47, International Monetary Fund, Washington, D.C. Henmming, R. (1998), 'Should public pensions be funded?' Working Paper no. 98/35, International Monetary Fund, Washington, D.C. - and Kay, J.A. (1982), 'The costs of the state earnings-related pension scheme', EconczncJoural, vol. 92, June, pp. 300-319. 38 Holzmann, R. (1998a), 'A World Bank perspective on pension reform', Pension Reform Priner series, Social Protection Discussion paper no. 9807, World Bank, Washington, D.C. - (1998b), 'Financing the transition to multipillar', Social Protection Discussion Paper no. 9809, World Bank, Washington, D.C. Hurd, M.D. and Shoven, J.B. (1985), The distributional impact of social security', 'in Wise, D.A. (ed.), Pensiozs, Labor and bxdiviI Choice, University of Chicago Press. Iglesias, A. and Palacios, R-J. (1999), 'Managing public pension reserves: evidence from the international experience', Pension Reform Primer series, Social Protection Discussion Paper, World Bank, Washington, D.C., forthcoming. Johnson, P.G. (1999), Older Geing Wuer, Institute of Chartered Accountants in Australia, Sydney (also available from Institute for Fiscal Studies, London). Kalisch, D.W. and Aman, T. (1998), 'Retirement income systems: the reform process across OECD countries', Ageing Working Paper no. 3.4, OECD, Paris. Kotlikoff, L.J. (1992), Gezeratnl AcwawiliV KnoUig Wi4o Pays, and WN, for W1w We Spend, Free Press, New York. -, Smetters, K. and Walliser, J. (1997), 'Opting out of social security and adverse selection', Working Paper no. 6430, National Bureau of Econornic Research, Cambridge, Mass. Lee, R and Skinner, J. (1999), Will aging baby boomers bust the Federal budget?' Jownal of EccnXnc PerypetiLs, vol. 13, no. 1, pp. 117-140. McHale, J. (1999), 'The risk of social security benefit rule changes: some international evidence', Working paper no. 7031, National Bureau of Economic Research, Cambridge, Mass. Miles, D. (1998), The reform of pensions systems in the UK and Germany: Winners and losers across the generations', mi7w, Imperial College, London. - (1999), Modelling the impact of demographic change on the economy', Econwc Jounal, vol. 109, January, pp. 1-36. OECD (1988), RefonngPuic Pesions, Paris. - (1995), The Trositixnfiran Work to Retii4 Social Policy Studies no. 16, Paris. - (1996), Agem7g in OECD Cao&tdes. A Critial Policy CYial. Social Policy Studies no. 20, Paris. - (1998a), Ma Prosperity in an Ageing Socity, Paris. - (1998b), 7he Tax-Be* Position of Emap1qes, Paris. Palacios, RJ. (1998), 'A note on diversification between funded and PAYG schemes', rnzrn, World Bank, Washington, D.C. - and Rocha, R. (1998), 'The Hungarian pension system in transition', Pension Reform Primer series, Social Protection Discussion Paper no. 9805, World Bank, Washington, D.C. - and Whitehouse, E.R. (1998), 'Individual choice and the transition to a funded pension system', Pension Reform Primer series, Social Protection Discussion Paper no. 9812, World Bank, Washington, D.C. 39 Palmer, E. (1999), The Swedish pension reform - framework and issues', mniw., National Social Insurance Board, Stockholm. Queisser, M. (1998), The S nd-Gmeratin Pension Refoms in Latin Amrrica, Ageing Working Paper no. 5.4, OECD, Paris. Roseveare, D., Leibfritz, W., Fore, D. and Wurzel, E. (1996), 'Ageing populations, pension systems and government budgets: simulations from 20 countries', Economics Department Working Paper no. 168, OECD, Paris. Samuelson, PA. (1958), 'An exact consumption-loan model of interest with or without the social contrivance of money', JoumalfPoliial Eccwxny, vol. 66, pp. 467-482. Stjohn, S. (1998), Pension provision in New Zealand', paper presented to pension reform conference, Institute for Fiscal Studies, London, March. Sunden, A. (1998), 'The Swedish pension reform', Znwm. Tanner, S. (1998), 'The dynamics of male retirement behaviour', Fiscal Studies, vol. 19, no. 2, pp. 175-196. United Kingdom, Department of Social Security (1984), 'Population, pension costs and pensioners' incomes', London. - (1997), 'Guaranteed secure pensions for all, says Peter Lilley', Press Notice no. 97/044, London. - (1998), A New CntrVactfor Welfanm ParJp in Pensios, Cm 4179, London. Vaupel, J.W. and Lundstrom, H. (1996), The future of mortality at older ages in developed countries', in Lutz, W. (ed.), 7he Fure Populaion of th Wote d W4 Ca We Assume Today? Earthscan Publications, London. Walliser, J. (1997), 'Understanding adverse selection in the annuities market and the impact of privatising social security', n., Congressional Budget Office, Washington, D.C. - (1998), 'Privatising social security while limiting adverse selection in annuities markets', mzb7w, Congressional Budget Office, Washington, D.C. Whitehouse, E-R. (1998), 'Pension reform in Britain', Pension Reform Primer series, Social Protection Discussion Paper no. 9810, World Bank, Washington, D.C. - (1999), 'Pension reform, consumer financial literacy and public information: a case study of the United Kingdom', Pension Reform Primer series, Social Protection Discussion Paper, World Bank, Washington, D.C., forthcoming. - and Wolf, M. (1997), 'State retirement plans', Fiz'iial Tnes, March. World Bank (1994), Atirg the OldAge Cnsis: Polxies to Protet dx Old dne Giuthd, Oxdord University Press. 40 Summary Findings Public pension programmes in OECD countries are in difficulties. With ageing populations, and declining participation of working age men in paid work, existing pension arrangements are likely to be unsustainable in the future in many of the richer OECD countries. Indeed, supporting existing pension commitments, even before the 'baby boom' generation reaches retirement, has already proved problematic in countries such as Italy. Some governments have already taken steps to tackle the pension issue but there is inevitably conflict over who will bear the burden of retrenchment: will it be current taxpayers, current pensioners, or future generations of taxpayers and pensioners, perhaps not yet born? This paper considers several issues. It examines the evidence as to whether public pension programmes in some richer OECD countries are indeed in need of major surgery, focusing in particular on the issue of fiscal sustainability. It then considers why programmes have got into financial difficulties. Consideration of this issue provides some clues as to what type of reform process is likely to be viable and credible. The paper then examines the strengths and weaknesses of some reform strategies. A central issue considered there is whether pension programmes should be funded or unfunded. ) ~~~HUMAN DEVELOPMENT NETWORK